RNAi Screening: New Approaches, Understandings and Organisms

Abstract

RNA interference (RNAi) leads to sequence-specific knockdown of gene function. The approach can be used in large-scale screens to interrogate function in various model organisms and an increasing number of other species. Genome-scale RNAi screens are routinely performed in cultured or primary cells or in vivo in organisms such as C. elegans. High-throughput RNAi screening is benefitting from the development of sophisticated new instrumentation and software tools for collecting and analyzing data, including high-content image data. The results of large-scale RNAi screens have already proved useful, leading to new understandings of gene function relevant to topics such as infection, cancer, obesity and aging. Nevertheless, important caveats apply and should be taken into consideration when developing or interpreting RNAi screens. Some level of false discovery is inherent to high-throughput approaches and specific to RNAi screens, false discovery due to off-target effects (OTEs) of RNAi reagents remains a problem. The need to improve our ability to use RNAi to elucidate gene function at large scale and in additional systems continues to be addressed through improved RNAi library design, development of innovative computational and analysis tools and other approaches.

Introduction

RNA interference (RNAi) is a conserved endogenous activity¹ that can be harnessed as a tool for functional genomics studies^2-8. With RNAi, gene-specific reagents are introduced into cells, triggering “knockdown” or reduction of gene function via sequence-specific degradation and translational interference of mRNA transcripts. RNAi screening provides a powerful reverse-genetic approach to large-scale functional analysis in cultured cells and in an increasing number of in vivo systems. Like genetic screening, RNAi screening allows for identification of genes relevant to a given pathway, structure or function via association of a mutant phenotype with gene knockdown. Like chemical screening, RNAi screening is amenable to miniaturization and automation, facilitating high-throughput studies. Due at least in part to the ease of delivery of RNAi reagents and resources available, C. elegans, Drosophila cells and mammalian cells have been the most-used systems for RNAi screening. Indeed, screens in these systems have already led to important new insights into a wide variety topics, including infectious disease, cancer, signaling and aging^{2, 3, 6, 8-16}. Moreover, RNAi screening has benefitted from input from a variety of other fields, in particular engineering and computer science, for example to improve methods for automated high-content image acquisition and analysis¹⁷.

Over the years, researchers have gained a better understanding of best practices for RNAi screening, both through performing screens and through study of endogenous RNAi pathways. In particular, recent improvements and refinements in methods for in vivo RNAi screening in Drosophila and mice have opened the doors to an increasing number of large-scale in vivo studies in those systems³. RNAi has been evolutionarily conserved and thus, it is being used to study an increasing number species for which functional genomics would otherwise not be feasible^{3, 18-22}. Despite all this progress, however, the problem of off-target effects and other sources of false discovery remain ongoing challenges. Improvements in reagent design, reagent delivery, assay design and data analysis have increased the quality of RNAi screen results in recent years. However, the picture remains complex in terms of understanding and addressing all possible sources of false positive and false negative results^{23, 24}. Despite these caveats, RNAi screening remains a powerful method-of-choice for genome-scale interrogation of gene function in an increasing number of systems, and the results of RNAi screens continue to provide new insights into diverse topics in biology and biomedicine. Below, we provide an overview of RNAi screening in cells and in vivo, focusing on new developments and results, as well as innovations stemming from interaction with other fields of study.

RNAi SCREENING IN CELLS

Why screen in cultured cells?

RNAi technology opened the doors to performing functional genomics in human cells and other types of cultured and primary cells. Cell-based RNAi screening builds upon established instrumentation, assays and other methods previously developed for chemical screening in cells. Overall, cell-based RNAi screening provides a relatively rapid and accessible platform for genome-scale functional studies^{2, 4, 5, 7}. A large number of RNAi screens has been performed in Drosophila and mammalian cultured cells². More recently, researchers have developed methods for screening neuronal and muscle primary cells derived from dissociated Drosophila embryos^25-27, as well as primary Drosophila haemocytes²⁸. In addition, an increasing number of studies are being performed using mammalian stem cells (reviewed in ^{29, 30}). The availability of transcriptome data for tissues, tumors and cell lines, made possible by next-generation sequencing technologies, is likely to shape choices and interpretation of cell-based RNAi screen data^{5, 15, 23, 31}. For example, transcriptome data may help us to understand the extent to which networks present in a cell line reflect what is happening in vivo, and detection of single-nucleotide polymorphisms (SNPs) can reveal mismatches to reagents that are relevant to the interpretation of results.

Reagent Libraries for Cell-Based Screening in Drosophila and Mammalian Cells

RNAi screening relies on the availability of genome-wide or other large-scale RNAi reagent libraries, with one or more unique RNAi reagent directed against each target gene. The appropriate RNAi reagent library for cell-based screening depends upon the cell type, approach and method of reagent delivery⁷. In Drosophila cells, the lack of an interferon response and ability of most cell types to take up the reagent in solution makes it possible to use in vitro synthesized long double-stranded RNA (dsRNAs) as the RNAi reagents². Reagents in the form of small interfering RNAs (siRNAs), endoribonuclease-prepared siRNAs (esiRNAs) or small hairpin RNAs (shRNAs) are typical for mammalian cell screens^{2, 4, 7, 32-34}, as long dsRNAs can evoke non-specific cellular responses that interfere with cell-based assays.

Design of effective and on-target RNAi reagents remains an ongoing challenge^{23, 24}. Useful tools for evaluation of RNAi libraries include NEXT-RNAi (http://www.nextrnai.org/)³⁵. A new approach was used recently to identify RNAi reagents conferring robust knockdown in mammalian cells³⁶. The results of their analysis of 20,000 RNAi reporters suggest that shRNA reagents conferring robust knockdown are relatively rare and may help provide new insights into effective reagent design. Improved access to information about reagent designs and results may contribute to the ability to learn new rules for effective design in the future. In addition to sequence-based efforts to improve RNAi reagent design, some researchers are working to achieve robust knockdown by combining RNAi with other approaches, such as U1 interference³⁷. Others are exploring the effects of adding various chemical modifications to siRNAs with the goal of developing more effective reagents^{24, 38}.

Arrayed Screening in Cultured Cells

With an arrayed screen, each RNAi reagent (or mini-pool of reagents, such as a set of independent siRNAs directed against a single gene) is contained in a separate well of a micro-well plate, such as a 96- or 384-well plate. Thus, following the experiment, the identity of the reagent can be determined by checking a database or spreadsheet that tracks which reagents were present in which wells. Researchers are using several different types of assays with arrayed screening approaches, facilitated by a number of different types of assay readout instruments (see below). Further miniaturization of arrayed screens has been achieved through the use of microarray slides on which the siRNAs or other RNAi reagents have been printed, facilitating reverse transfection of reagents into cells³⁹. Recent applications of this approach using Drosophila or mammalian cells have looked at signal transduction⁴⁰, differentiation⁴¹ and host-pathogen interactions⁴². The next frontier in arrayed screening may be the use of specially designed microfluidics instruments and micro-well platforms to facilitate single cell analyses^{43, 44}.

Assay Readouts for Arrayed Screening

Several types of cell-based assays are made possible by arrayed-format screening. For example, researchers can measure ATP levels, transcriptional activity and protein stability using “plate-readers” (i.e. luminometers and fluorimeters) that measure whole-well intensity of luciferase or fluorescence readouts. To date, a large number of plate-reader screens have been performed in mammalian and Drosophila culture cells². As for other screening approaches, arrayed plate-reader screens can be combined with addition of a treatment such as a drug, infectious pathogen, or environmental stress in order to identify modifiers of the phenotype normally induced upon exposure to that treatment (see for example ^45-48). Conversely, researchers are also sensitizing cells using RNAi treatment against one gene, followed by chemical screening⁴⁹. A new approach known as real-time cell analysis (RTCA) has facilitated time-lapse screening⁵⁰.

Resolution at the individual cell level is essential for some screen assays. Fluorescence-assisted cell sorting (FACS) has been used in arrayed-based screens to determine DNA content or the relative levels of two markers at the individual cell level (see for example ^51-53). However, due to the relatively slow speed of FACS analysis in micro-well formats as compared with some imaging approaches, researchers more often turn to imaging for cellular and sub-cellular resolution of various fluorescent-labeled dyes, probes or antibodies. Screen imaging instruments typically balance acquisition speed against image quality or resolution ¹⁷. Relatively rapid, whole-well imaging can be achieved using instruments such as laser scanning cytometers. Alternatively, sub-cellular resolution can be achieved using epifluorescence or confocal screening microscopy.

High-Content Image-Based Assays

The availability of high-throughput, high-content imaging instruments has made it possible to obtain simultaneous readouts of various different fluorescent or other visual markers. The sheer number and volume of images obtained in high-content image-base screening require the use of automated solutions to identifying the sub-set of RNAi reagents that generate in the phenotype(s) of interest. Using image analysis software tools such as CellProfiler⁵⁴, hundreds of different features can be extracted from screen image datasets and then used to define or identify phenotypes that are relevant to the topic being addressed (see Sidebar, High-Throughput, High-Content Imaging in RNAi Screening). The state-of-the-art in image-based assays involves live cell and time-lapse screening (see for example ^{40, 55-58}). Although time-lapse screen imaging with live cells opens the door to new types of studies, these approaches also add significantly to acquisition time, increase the total volume of data that must be managed and stored, and add to the burden of image data processing and analysis. This problem has begun to be addressed with analysis tools such as CellCognition⁵⁹ and the development of start-to-finish automated platforms for sample processing, imaging and analysis⁵⁷.

Pooled Screening in Cultured Cells

Pooled screening provides a convenient method for screening large RNAi reagent collections, such as genome-wide mammalian shRNA libraries^{2, 32}. With a pooled screen, the RNAi library is introduced into cells at random by DNA transfection or perhaps more commonly, by viral transduction, with the goal of introducing one RNAi reagent per cell. Pooled screen readouts depend upon comparison of two or more populations of cells and have contributed to our understanding of a number of topics, including cancer^{2, 32}. A related approach is to introduce library-transduced pools of cells into mice to perform what are known as ex vivo screens (see below, in vivo Screening in Mice).

Deconvolution of the results of a pooled screen—i.e. identifying the sub-set of RNAi reagents that are enriched and/or depleted in the experimental versus starting or control pools—is a key step. Deconvolution of pooled screen results is typically done using microarrays or next-generation sequencing to detect the total population of reagents present in each pool. Due to the technical challenge of deconvolving results obtained from very complex pools, researchers divide genome-scale collections into smaller pools, e.g. six pools of ~13,000 unique reagents^{74, 75}. A newly developed microarray-based resource reportedly allows for deconvolution of pooled shRNA screens with up to ~90,000 unique shRNAs⁷⁶.

Double-Knockdown Screens

Another area of innovation in cell-based, arrayed screening is large-scale combined targeting of more than one gene, such as for detection of synthetic lethal interactions⁷⁷. As we have learned from studies in yeast, combinatorial approaches can be particularly powerful in addressing issues of redundancy in genetic networks⁷⁸. For Drosophila cell-based screens, RNAi reagents directed against two different genes can simply be combined in solution. Large-scale pairwise RNAi screens in Drosophila cells have provided insights into redundancy and connectivity of conserved signal transduction pathways^{79, 80}. One of these studies included rigorous analysis of all pairwise combinations among a large number of genes. Among other findings, the study showed that combinatorial RNAi can reveal results that could not have been predicted based on single gene analyses⁷⁹. An alternative approach to reagent delivery for Drosophila cell-based or in vivo assays is expression of shRNAs targeting two different genes via a single transcript⁸¹. Researchers are exploring similar plasmid-based approaches, including expression of multiple hairpins from a single expression cassette, in mammalian cells⁸². Because of the large number of possible gene combinations, performing combinatorial RNAi screens in miniaturized formats such as using microarray slides might facilitate this type of screen.

RNAi SCREENING IN VIVO

Why screen with RNAi in vivo?

Many complex phenotypes cannot be reduced to a cell-based assay, thus requiring gene function to be directly analyzed in vivo. RNAi screening in vivo provides a relatively fast and straightforward route for screening a phenotype of interest in a tissue and stage-specific manner. Additionally, in vivo RNAi makes functional genomics studies possible in organisms for which classical genetic approaches have not been developed but for which genome or transcriptome sequences have become available—as gene annotations are necessary for the design of RNAi reagents ³. Species for which RNAi-based approaches and libraries are now being developed include Lepidoptera¹⁸ and other insects²²; many types of ticks⁸³; Hydra⁸⁴; planarians⁸⁵; a variety of plants^{20, 86}; and pathogens such as Trypanosomes ^{87, 88} (See Sidebar, “RNAi as Tool and Treatment in Disease Vectors, Parasites and Pests”).

Due to species-specific differences in reagent uptake and endogenous RNAi pathways, what specific RNAi reagents and methods of delivery are most appropriate must be worked out for each species, with common methods including dsRNA injection and feeding^{19, 83, 97}. Moreover, RNAi knockdown is systemic in some species—spreading from cell to cell—but not in others²². For the crop pest Bactrocera dorsalis (oriental fruit fly), one group observed knockdown of target genes upon short exposure to RNAi treatments and up-regulation of the genes following prolonged treatment ⁹⁶, emphasizing the need to carefully test the effects of RNAi treatments in each new species under study. The Lepidoptera community provides a nice example of how researchers can collaborate to try to improve RNAi methodologies in a related group of organisms, i.e. by sharing information and working to establish appropriate controls¹⁸ (see http://insectacentral.org/RNAi).

In vivo RNAi Screening in C. elegans

RNAi was first identified and characterized in C. elegans, and many in vivo RNAi screens have now been performed, leading to new understandings in diverse biological and biomedical topics^{3, 11}. In C. elegans, RNAi is systemic and heritable. Notable recent examples of genome-wide C. elegans RNAi screens include studies of aging and obesity^{13, 98-101}. Although methods for RNAi screening in C. elegans are well established, screening in this system is not without caveats. For example, a recent study of the widely-used Ahringer feeding library suggests that a significant number of gene annotations attached to strains in the collection need to be updated, and that some of the bacterial strains in the collection do not express dsRNAs or express dsRNAs that do not correspond to C. elegans genes¹⁰².

Similar to cell-based screening, RNAi screening in C. elegans increasingly relies on the use of image-based and modifier screens. In one recent screen, for example, researchers fed the transparent nematodes vital lipid dyes and used stimulated Raman scattering (SRS) microscopy to visualize stored lipid droplets^{99, 100}. A sophisticated high-content imaging and analysis approach was recently used to describe a global network of essential genes⁷⁰. A recent study of oxysterol-binding protein-related proteins is notable in that it was conducted as an enhancer screen in a quadruple-mutant background¹⁰³. Another enhancer screen demonstrated that genes with redundant or partially redundant functions can be revealed using RNAi screening in sensitized backgrounds¹⁰⁴. Moreover, although in C. elegans RNAi knockdown is less efficient in the nervous system than in other tissues, the use of sensitized backgrounds can facilitate screening for neuronal phenotypes¹⁰⁵.

In vivo RNAi Screening in Drosophila

With a well-annotated genome and wealth of other molecular genetic tools behind it, Drosophila melanogaster is another popular model system for in vivo RNAi-based screens. Although there is a report of systemic spread of RNAi in vivo in Drosophila as a mechanism for antiviral immunity¹⁰⁶, RNAi knockdown induced via injection or expression of dsRNAs acts cell-autonomously in Drosophila, facilitating tissue- and stage-specific studies ^{3, 107}. Expression of long or short dsRNA hairpins via a transgene is a flexible and robust option. Thus, it has become the method-of-choice for RNAi in this species, and three groups have built genome-scale libraries for RNAi screening in Drosophila (reviewed in ³).

The ability to induce RNAi in specific tissues and stages opens the door to screening not just in embryonic or larval stage animals but also in adults, even when knockdown in early stages is associated with lethality. Moreover, using shRNAs rather than long dsRNAs have made it possible to achieve robust knockdown not only in somatic cells but also in the germline¹⁰⁸. Recently reported genome-wide RNAi screens in vivo in Drosophila are notable in focusing on medically-relevant topics, including pain perception, obesity, heart function, bacterial infections of the gut, neural stem cell self-renewal, and neurological disease^109-114. Drosophila RNAi has also recently been used for relatively rapid in vivo follow up on ~500 gene candidates identified using a cell-based screen for regulators of Notch signaling¹¹⁵.

In vivo RNAi Screening in Mice

In addition to performing pooled screens by comparing differently treated cell populations grown in culture, researchers have also introduced pools of shRNA-transfected cells into mice, an approach referred to as ex vivo screening. This approach combines the relative ease of introducing large-scale libraries into a pool of cultured cells with the advantages of placing cells in an in vivo context. The ex vivo screening approach has proved particularly useful for cancer-based studies, in which transduced cells can be assayed for their ability to contribute to tumor formation following introduction into the host animal (see for example ^{116, 117}).

RNAi screening in vivo directly in mice, such as via inducible expression of transgenes, is in early stages as compared with in vivo screening with Drosophila or C. elegans. Nevertheless, recent breakthroughs in reagent design and delivery suggest that high-throughput RNAi screening will soon be feasible in at least some cell types and tissues. Reports from S. Lowe and colleagues describe inducible constructs that can be introduced into embryonic stem cells (ES) cells, facilitating RNAi in ES cells or production of transgenic mice for in vivo RNAi^{118, 119}. An alternative approach based on infection with lentivirus facilitates in vivo RNAi in accessible tissues such as skin^{120, 121}. Lentiviral vectors specifically designed with in vivo approaches in mind have also been reported and should help facilitate in vivo approaches in mice in the future¹²².

OFF-TARGET EFFECTS IN RNAi SCREENS

The problem of false discovery in RNAi screens—i.e. false positive and false negative screen results—is made particularly clear by the results of meta-analyses of multiple related screens in Drosophila or mammalian cells, which reveal poor reproducibility between or among related screens^{123, 124}. The lack of overlap appears to be due to both false positive and false negative results. Many sources of false discovery are inherent to high-throughput studies^{4, 7, 23, 125}. Problems like instrument errors or flawed assay designs can often be detected and addressed during assay development and optimization. Statistical noise is inherent to any large-scale study but can be kept to a minimum through conducting an appropriate number of replicate tests and applying appropriate statistical analyses (see below, Limiting False Discovery). For image-based screens, image processing such as correcting for uneven illumination can be done prior to image analysis. Even after these sources of false discovery are addressed, however, problems of false discovery due to RNAi reagent design remain.

Contribution of RNAi Reagents to False Discovery

False positive results attributable to RNAi reagents include both sequence-independent and sequence-specific effects. Sequence-independent effects include invoking an interferon response and toxicity of the reagent delivery method^{125, 126}, as well as general disruption of the endogenous miRNA pathway ¹²⁷. Sequence-independent effects can often be addressed by using reagent and delivery methods that are appropriate to the specific cell type or organism being used in the study⁷. A recent study of Drosophila cell screens underscores the idea that choosing an appropriate number of reagents and replicates is important to limiting false negative results³¹.

Arguably the more challenging goal in effective reagent design has been to address sequence-specific false positive results, or OTEs. These are due to sequence-specific recognition of transcripts other than the intended target by the RNAi reagent, followed by entry into the RNAi or miRNA pathways. To limit sequence-specific OTEs, most design algorithms avoid regions in the target sequences that have 19 or more base pairs of contiguous nucleotide identity to another mature transcript, as a 19-mer is sufficient to induce RNAi knockdown of a target transcript. Application of the “19 based-pair rule” has resulted in a marked improvement in RNAi reagent libraries for mammalian and Drosophila cells as compared with early libraries¹²⁸. Nevertheless, shorter perfect matches and imperfect matches to other mRNA sequences also contribute to OTEs^{23, 129}. In addition, a recent report suggests that at least in Drosophila, matches to intronic sequences might be relevant as well¹³⁰. The same researchers have also shown that unique reagents against a gene can share common OTEs¹³¹. In the case of imperfect matches, the RNAi reagents appear to shuttle transcripts into the endogenous miRNA pathway. Consistent with this, analysis of the of the hits from one recent screen revealed that a large number of reagents that scored as positive in the primary screen were acting via miRNA-like effects on transcripts that encode TGF-beta receptors¹³². This allowed the researchers to identify endogenous miRNAs that might be involved in TGF-beta signaling. Nevertheless, the initial goal of uncovering new protein-coding genes involved in TGF-beta signaling was not achieved, and this serves as an important caution for analysis and interpretation of similar screen results.

RNAi reagent design also contributes to false negative results, as some reagents do not result in robust knockdown of the target gene. In Drosophila one way to improve the efficiency of long dsRNAs is to co-express Dicer2 and the RNAi reagent¹³³, as over-expression of Dicer2 presumably improves the processing of the dsRNAs into siRNAs. Screening in sensitized strains to enhance the effects of RNAi is also done in C. elegans, such as for screens of the nervous system, wherein the feeding approach is less effective⁴. In the case of transgenic or viral delivery methods, robust expression of the active RNAi reagent can be critical to achieving robust knockdown, such that optimization of vector designs, delivery, chromosomal insertion sites, etc. can help reduce false negative discovery. The results of a recent study using a fluorescent protein sensor to detect effective shRNA sequences suggest that more can be learned about rules for effective RNAi reagent design, such that development of new and improved RNAi reagent libraries can be expected to continue³⁶.

Limiting False Discovery

Prior to conducting a screen, it is important to establish what number unique RNAi reagents per gene and what number of replicate tests will yield meaningful data, as well as carefully assessing pilot screen data to check for instrument error or other sources contributing to false discovery^{4, 7, 31, 134, 135}. Subsequent to conducting a screen, a number of computational and experimental approaches can be used to limit false discovery. A set of best practice guidelines for statistical analysis of mammalian siRNA arrayed screen data has been put forward¹³⁵. Additional statistical approaches specifically designed for RNAi reagents such as siRNAs have also been proposed^{136, 137}. Cut-off values can be chosen strictly based on analysis of the screen data or can follow an informed approach, such as using information about protein-protein interactions or functional networks to establish an appropriate statistical cut-off value for a particular screen ¹³⁸. After screen data are published, other groups should be able to reanalyze datasets to test the utility of new statistical approaches. This would be facilitated by general adoption of recently proposed minimal information about RNAi experiments (MIARE) standards (see http://miare.sourceforge.net/HomePage) and deposition of data into a centralized database such as NCBI PubChem (http://pubchem.ncbi.nlm.nih.gov/).

Other methods for limiting false discovery are appropriate for specific screen assays, cell types, topics, or approaches. For example, recent reports suggest that careful analysis of multiple parameters extracted from high-content image data can point to the sub-set of parameters that are most informative, limiting false positive results^{64, 65}. For some cells or systems, it is also possible to remove false positive results by comparing screen hits with transcriptome data. In this case, researchers make the underlying assumption that screen hits corresponding to genes known to be expressed in a given cell type or tissue are more likely to represent on-target true positive results than hits corresponding to genes for which there is no evidence that the gene is expressed³¹. Comparison with pathways or networks culled from the published literature or large-scale proteomics datasets is also proving to be an appropriate method for limiting false discovery¹³⁹. In these cases, false positive results can be limited by excluding screen hits that are not supported in the orthogonal dataset, and potential false negative results can be addressed by adding genes to the list for secondary analyses, such as genes that did not show up as strong positives but are components of a given pathway or complex¹⁴⁰.

Testing of two or more non-overlapping RNAi reagents per gene is a general standard for initial verification of primary screen results. Increased confidence in cell-based RNAi results can also be achieved by testing for comparable effects in vivo in the same species or testing for comparable effects in cells or in vivo in another species (see for example ^{115, 119, 141}). Ultimately, the ‘gold standard’ test for an on-target effect is rescue, such as with a genomic fragment, cDNA or open reading frame construct designed to evade RNAi ^142-150, and confidence in screen data is further increased when results are confirmed using other molecular genetic methods.

Conclusion

RNAi has proven to be a powerful tool for systematic testing of gene function, including at genome-wide scale^2-8. Indeed, the results of RNAi screens have already led to new understandings of gene functions and networks in the context of basic cell biology and biomedicine^{2, 5, 6}. Topics under particularly intense study using RNAi screening include cancer biology and resistance to anti-cancer treatments, interactions between host cells and viral or bacterial pathogens, and basic cellular functions such as growth, division and metabolism^{2, 6, 8-11, 14, 16}. Pooled screens, including ex vivo screens in mice, are having particular impact in understanding cancer^{32, 116, 117}. Arrayed cell-based screens have also proved informative and recently, they have gained from the availability of increasingly sophisticated assay read-out instruments, such as for high-content imaging, as well as increasingly sophisticated analysis tools^{4, 17, 62}.

Large-scale in vivo RNAi in model systems has led to new insights into topics like obesity and aging, and, making the most of what can be done in vivo but not in cells, studies of complex behaviors such as nociception^{3, 98, 99, 101, 112}. RNAi requires little more than transcript annotations, a reagent library, and a method of delivery of reagents and consequently, genome-scale in vivo RNAi screens are becoming possible in an increasing number of health-relevant species³. Improvements in the speed at which we can annotate transcripts—i.e. using next generation sequencing technologies—and the ease of RNAi reagent delivery in some systems suggest that we can expect rapid development of RNAi tools for an increasingly broad spectrum of organisms⁹³. The many known and emerging caveats to interpretation of RNAi screen results, including but not limited to the enduring problem of OTEs²³, should be better addressed as new reagent libraries, approaches and instrumentation are developed for established models and emerging organisms. Moreover, as our understanding of the biology behind RNAi effects continues to grow, we can continue to exploit that knowledge in the design of more robust and specific reagents, approaches and analysis tools. Thus, both the scope and the quality of results from RNAi screens can be expected to improve in the future.

Challenges and Strategies for High-Content Imaging in RNAi Screening.

Challenges for image-based screening include deciding on an instrument, storing and managing large datasets, and determining which features or parameters are most informative^{17, 60-62}. Deciding what to image can also affect results; for example, in a host-pathogen screen, the results differed when researchers imaged the virus versus imaging of host cell features indicative of infection⁶³. Image analysts and statisticians help researchers make the most of screen image data, including through development of improved and new software tools such as Cell Profiler and CellCognition^{54, 59}. An approach that is growing in popularity and utility is multi-parametric image analysis—i.e. measurement of many different features, such as signal intensity, size, shape and/or texture—which is sometimes done in combination with machine learning. Multi-parametric analysis and machine learning are being used not just to identify pre-defined phenotypes but also to identify new phenotypes, limit false discovery, and place genes in networks based on shared phenotypes^{57, 64-72}. Determining which of the many available image analysis software tools might be appropriate for analysis of a specific dataset is another challenge. Establishment of benchmarking principles may help researchers not just to identify an appropriate analysis approach but also to tailor image acquisition parameters to best fit their downstream goals⁶⁰. Efforts to standardize how we identify and describe sub-cellular features⁷³ may also have an impact on high-content screening, as standardizing terms can help facilitate cross-comparison of results from multiple screens.

RNAi as Tool and Treatment for Disease Vectors, Parasites and Pests.

RNAi makes it possible to do functional studies in species for which other genetic tools are not available³. Thus, it is now possible to study gene function directly in organisms that impact human health, including disease vectors (e.g. mosquitoes), parasites, pathogens, and crop pests^{19, 21, 22, 83, 89, 90}. Topics of particular interest for RNAi screening in health-relevant species include viability, fertility, innate immunity and biocide resistance^{87, 88, 91, 92}. Improved genome-wide libraries were used recently in RNAi screens of the bloodborne pathogen that cases African trypanosomiasis (or sleeping sickness), Trypanosoma brucei ^{87, 88}. A method called reciprocal allele-specific RNAi has been used to study variability in the ability of Anopheles mosquitoes to transmit the malaria parasite Plasmodium⁹². For species lacking gene annotations, next-generation sequencing is being used to perform large-scale identification of mRNA transcripts, followed by RNAi reagent design and production⁹³. In addition to its use as a research tool, RNAi is also being explored as a method for pest population control, and in the case of disease vectors, for controlling infection of the vectors by disease agents^{19, 22, 90}. Researchers are also testing expression of RNAi reagents in crop plants as a defense against infection^{94, 95}. Issues such as off-target effects, differences in endogenous RNAi pathways and activities, genetic variation within a species, a report of up-regulation of gene activity upon prolonged exposure to RNAi, and the potential of RNAi treatments to affect other species serve as important cautions as research efforts in these areas move forward^{95, 96}.